how to use Lux.jl utility 'BinaryCrossEntropy'

[CMoebus]

Hi,
I am am new to Lux.jl and want to use the 'BinaryCrossEntropy' utility according to the small example (https://lux.csail.mit.edu/stable/api/Lux/utilities#Lux.BinaryCrossEntropyLoss):

julia> bce = BinaryCrossEntropyLoss();
julia> y_bin = Bool[1, 0, 1];
julia> y_model = Float32[2, -1, pi]
3-element Vector{Float32}:
  2.0
 -1.0
  3.1415927

but my Pluto/Julia script:

begin 
	using Lux, LossFunctions
	LossFunctions.CrossEntropyLoss()                     # this was recognized
	bce = Lux.BinaryCrossEntropyLoss()                 # this not !
end # begin

led to this error:

 UndefVarError: `BinaryCrossEntropyLoss` not defined
    getproperty(::Module, ::Symbol)@Base.jl:31
    top-level scope@[Local: 4](http://localhost:1234/edit?id=7b7ed1be-57e9-11ef-3d33-2d8cf3743530#)

What was my fault ? Thanks for help, Claus

[avik-pal]

Can you share what is pkgversion(Lux), I am guessing an old version of Lux was installed for some reason.

[CMoebus]

v"0.5.17" https://uol.de/en/lcs/

…

________________________________ From: Avik Pal ***@***.***> Sent: Sunday, August 11, 2024 6:48:08 PM To: LuxDL/Lux.jl ***@***.***> Cc: Claus Möbus ***@***.***>; Author ***@***.***> Subject: Re: [LuxDL/Lux.jl] how to use Lux.jl utility 'BinaryCrossEntropy' (Discussion #842) ACHTUNG! Diese E-Mail kommt von Extern! WARNING! This email originated off-campus. Can you share what is pkgversion(Lux), I am guessing an old version of Lux was installed for some reason. — Reply to this email directly, view it on GitHub<#842 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ABOF6LTPCNO5VQGDLHQ4OIDZQ6IURAVCNFSM6AAAAABMK4TAAKVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTAMZQGY2DMMY>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[CMoebus]

Hi, I solved the problem by myself ! I deleted the Julia registry and let it rebuild again. Then I updated the Lux package. Now, I have the latest version pkgversion(Lux) = 0.5.63. 'BinaryCrossEntropy' is recognized now and the example in the doc is running.
Thanks for your patience, Claus